This article began, as so many articles do, with a vague question: What do the new polls from NBC and Marist mean about President Trump’s popularity and the 2018 midterm elections?

Those polls, released Wednesday night, show that Trump is very unpopular in three states that would seem to be critical to his reelection: Wisconsin, Michigan and Minnesota. He lost Minnesota narrowly in 2016, but his tiny margins in Wisconsin and Michigan were enough (combined with Pennsylvania) to give him the presidency. If Trump is viewed poorly in those three states, never mind picking up Minnesota, these don’t seem like numbers that will get him there.

Look at it this way. In each state, Trump’s support in 2016 as a percentage of the vote was much higher than his current approval ratings. This isn’t apples-to-apples; 2016 voting was Trump vs. the not-always-popular Hillary Clinton, not simply an assessment of Trump himself. But we know that approval ratings correspond to presidential reelection votes. (The figures for Michigan and Wisconsin are the same here, which is why there’s only one line.)

The net figures (Trump support minus Clinton support and approve minus disapprove) show the same downward trend.

In fact, that trend plays out in a slew of other states with recent approval-rating polling. We pulled polls from the past month or so and ran the same comparison.

The steepest drops from 2016 numbers to now were in three states critical to Trump: Ohio, Michigan and Wisconsin. In Ohio, a Marist poll showed Trump with a minus-8 approval rating a month ago, a low figure in a state that he won with a majority of the vote two years ago.

In only one state — New York — was there upward movement: The net margin of support for Trump in 2016 was lower than his current net approval. But Trump isn’t going to win New York, nor are Republicans going to do well there in November.

Which brings us back to the broader point.

We know that there’s a relationship between how Republicans are expected to fare in November and how Trump is viewed. The more popular Trump is, the better Republicans do on the generic ballot (that is, polling asking people whether they plan to vote for the Democrat or the Republican in the midterms). The opposite is also true.

Remember how I said this started with a vague question? Well, here’s where it ended up. I was curious about how that movement looked over time, so I plotted it out. Then I broke it out into months, creating 15 individual lines based on RealClearPolitics’s polling averages since May of last year, showing how the relationship between Trump’s net approval has compared to his party’s net position on the generic ballot.

Moving from bottom to top in each box below, Trump’s net approval (approve minus disapprove) is getting better. Moving from left to right, the net support for Republicans (respondents preferring a Republican member of Congress minus the percentage picking the Democrat) is getting better as well. Each line starts on the first of the month with that little square and wiggles around until the end of the month (or, for July 2018, the most recent data). The range on each axis is from zero to minus-20.

First, take it in.

Some things you’ll notice:

  • Trump’s net approval dropped quickly in May 2017 (the line moves down), but support for the GOP didn’t suffer.
  • In September, Trump’s net support rose and fell and rose again, again without affecting the generic ballot much. In part, this is a function of there not being many generic-ballot polls in that month, meaning the average didn’t move much.
  • In November, Trump’s net approval dropped, and so did net support for his party.
  • In December, Trump’s worst month of his presidency polling-wise, he ended about where he started, but the GOP’s position slipped.
  • In June of this year, his position and his party’s both dropped (moving down and to the left).
  • This month, there hasn’t been much movement at all.

Notice the general patterns here, though. Up-and-down and left-to-right line movement in a month; often, diagonal movement from lower left to upper right or vice versa. But not much movement from upper left to lower right — not much movement in which Trump’s numbers were headed one way and his party’s another.

This tells us a couple of things. It reinforces the link between the numbers, clearly, and it shows in a new way some trends we knew about. (Like that September was a good month for Trump, after the hurricanes on the Gulf of Mexico, and that December was very much not.)

What this doesn’t do, though, is tell us much about November. So if you’ve read this far expecting insight instead of interesting-looking graphs?

Well, I’m sorry.